Multivariate analysis of wind stress and curl over the Japan/East Sea, based on satellite scatterometry data

O. TrusenkovaPacific Oceanological Institute, Vladivostok, Russia

PICES 2010, October 22–31, 2010, Portland, Oregon, USA

MotivationMonthly mean wind stress and curl in the JES area

Stable conditions in winter, changeable and weak, on monthly time scale, wind in summer Are there any patterns in summer wind?

Leading complex EOF mode from wind vectors accounted for the general wind direction effect of the East Asia monsoon (Trusenkova et al., 2009).

Previous results from 1°-gridded NCEP reanalysis (ancillary data for the SeaWifS Project)

However, fraction of the total variance was not high any impact of higher modes?There were fictional alongshore wind gradient zones fictional curl zones.

Mode 1

Wind speed, m

Curl

To revisit the spatial/temporal patterns of wind stress and curl over the JES by multivariate statistical analysis of

satellite scatterometry data

Purpose

Sea winds from QuikSCAT

http://www.remss.com/qscat/scatterometer_data_daily

High resolution data over the water surface.

However, frequent gaps due to satellite swath divergence and rain contamination

QSCAT merged with reanalysis provides data over the sea and land.

Data

QSCAT/NCEP Blended Ocean Winds

from Colorado Research Associates,

July 1999 - July 2009,

34°-53°N, 127°-143°E (JES and adjacent land),

6 hours, 0.5°

grid,

1287 wind stress boxes and

386 stress curl boxes (over the sea only),

14724 times.

1) The complex form of the EOF analysis applied to wind vectors. X(r, t) = ∑

Ak*(r)Bk (t), where

X(r, t) = U(r, t) + iV(r, t), U/V are zonal/meridional wind components, Ak (r) = Ak (r)e-iϕ

are spatial CEOFs and * denotes complex conjugate, Bk (t) = Bk (t)e-iφ

are principal components (PCs), Ak , Bk are spatial and temporal amplitudes ~ mode intensity,ϕk , φk are spatial and temporal phases (-180°, 180°) ~ wind shear in space and time.

2) Correlations detection of low amplitude signals ~ wind over the landCovariances ~ signal magnitude anomalies from weaker signals can be lost.

3) Low-pass filtering by the inverse wavelet transform, Morlet mother wavelet of the 6th order.

Techniques of EOF analysis

• Original dataset. • Low-pass filtered dataset, with the 7- and 15-day cut-off periods.

Decompositions of wind stress vectors

Modes 1 – 3 from the original data and modes 1 & 2 from the filtered data pass the Montecarlo test and are non-degenerate in the sense of errors.

High fraction of variance (>50%) for mode 1 from the filtered data.

Similar variance fraction for mode 2. Breaks

CEOF modes of wind stress vectors: spatial patterns

2

Vectors for zero temporal phase°

Similar patterns of modes 1 & 2 from the original and filtered data.

CEOF modes of wind stress vectors: temporal patterns

Winter: strong wind Summer: weak wind

Winter: Persistent NW wind Summer: changeable wind

φ1 =135°

~ NW wind

Mode 1: general wind direction

Xk (r,t)=(Ak (r)e-iϕ(r))*Bk (t)e-iφ(t)

W NW N NE E SE S SW W

S, -90°

NW, 135°

E 0°

N, 90°

W ±180°

+ φφt =0°

~ eastern wind

PC 1, phase

Noisy original φ(t), patterns of seasonal wind shifts in the filtered data.

φ1 =0°

Mode 2: wind seesaw over the southern and northern JES

Typical contribution of mode 2 from the filtered data setφ = 0 ~ eastern wind

W NW N NE E SE S SW W

PC 2, phase

Mode 3: cyclonic / anticyclonic vortex (original fields only)

Typical contribution of mode 3

Cyclone Anticyclone

PC 3, phase

General wind direction: any contribution from the higher modes?

Mode 3 does not affect the wind shifts beyond the synoptic scale

Temporal amplitude: PC1/PC2=1.5.

Wind shift: < 10°

in the northern JES and <20°

off the Tsushima Strait

General wind directions over the JES can be deduced from the leading mode alone –

The East Asia Monsoon Mode

Wind shift due to CEOF 2

Monthly occurrence of the monsoon winds (from the temporal phase of mode 1 from low-pass filtered data)

Phase transform φ’ = -φ

+ 90°

N, 0°

+

φ’

φ’=0°

~ northern wind

φ’=315°

~ NW wind

5°-gradation histograms

S, 180°

E, 90°W, 270°

NW, 315°

Occ

urre

nce,

%

Monthly occurrence of the monsoon winds

NW: dominant in November through February;

W: separate mode in March and April and merged with NW in October; SW: occurs in April through September, strongest in May and June; Easterly modes: occur more frequently in late summer; NE: strong in August and September.

November – February: the only mode; April and September: fuzzy; July: bimodal (SW and E).

The southern wind is never a dominant mode.

Occ

urre

nce,

%

5°-gradation histograms

Seasonal wind shifts

Prevailing transitions: Spring: NW W SWSummer: SW SE EAutumn: E NE N NW or E SE SW W NW

Strong intraseasonal shifts beyond the synoptic time scale

W NW N NE E SE S SW W

PC 1, phase Winter

X1 (t) = (A1ave-iϕ1av)*B1 e-iφ1(t)

Autumn: E NE N NW

Autumn: E SE SW W NW

Intraseasonal variability beyond the synoptic scale (1 vector for 5 days shown)

Time series of monsoon winds (beyond the synoptic time scale)

Typical patterns of wind stress and curlJanuary NW October

Occ

urre

nce,

%

SW May E July

W March

NE September

Stresses averaged within histogram modes; curls computed from the average stresses.

Stresses from April through October smoothed with 3-point window.

Low-pass filtered wind stress curl fields, with the 40-day cut-off period.

Decomposition of wind stress curl

EV 2 and 3 are close, followed by a break

Eigenspectrum

break

Modes 1-3 are statistically significant

Mode 1 can be considered separately

Modes 2 and 3 are better considered together

EOF modes of low-pass filtered wind stress curl

×10-8

dyn/cm3

Winter mode

Cyclonic curl; curl dipole

off Vladivostok (Kawamura and

Wu, 1998), several dipoles near the western coast, AC curl zone off

Khabarovsky Krai coast,

cyclonic curl zone off the

East Korea Bay.

PC1 is above zero in November through February (mid March)

Topo

grap

hic

gaps

×10-8

dyn/cm3

Strong mode: winters 2000/2001, 2005/2006 Weak mode: winters 2001/2002 2006/2007, 2007/2008

The EKWC separation from the coast

January, stress from histogram

Warm season mode: combined effect of EOF 2 and EOF 3 (significant when PC 1 is close to zero)

Contribution of modes 2 and 3X1,2 (t) = A2 (r1,2 )B2 (t) + A3 (r1,2 )B3 (t)

Box 1 (central JES): EOF 2 dominates, corr(PC2, X1 ) = 0.98

Box 2 (northeastern JES): EOF 3 dominates, corr(PC3, X2 ) = 0.97

Curl (×10-8 dyn/cm3 )

Strong southward of 43°N

Strong northward of 43°N

Curl (×10-8 dyn/cm3

)

Warm season mode: the central JES

AC curl in late winter and autumn, oscillating between cyclonic and zero (slight AC) curl in late summer.

April 2000, 2006: strong cyclonic curl.

Eastern wind, frequent in late summer

NW wind in March and October: AC curl

Winter/summer curls for the median spatial EOFs; summer curl when exceeds the winter curl

October

March

July

April

April

Cyclonic curl in spring/early summer and autumn, frequent AC curl in late summer.

Western or SW wind, frequent in spring and autumn

Warm season mode: the northeastern JES

Winter/summer curls for the median spatial EOFs; summer curl when exceeds the winter curl

October

July

• The East Asia Monsoon pattern with characteristic wind directions and seasonal shifts is the leading complex EOF mode of wind stress over the JES.

•Contribution of higher modes is not significant.

•The results are consistent with the previous findings from reanalysis data (Trusenkova et al., 2009).

•Winter pattern is the leading EOF mode of wind stress curl. Fine features are resolved, such as several curl dipoles related to orographic gaps.

•Over the central JES, the AC wind stress curl prevails in late winter and autumn, while oscillations between the cyclonic and weak AC curl occur in summer.

•Over the northeastern JES cyclonic curl prevails in spring and autumn.

Conclusion

Thank you!